水文时间序列趋势和跳跃分析的再抽样方法研究

Trend and Change Point Analysis of Hydrological Time Series Based on Resampling Methods

用再抽样方法估计水文时间序列趋势或跳跃的显著性水平,可通过分块的方法消除水文时间序列自相关性的干扰,具有直接、精确的优点。介绍了再抽样方法在水文时间序列趋势和跳跃分析中的原理与方法;针对黄河三门峡水文站1919～1989年实测年径流量资料,采用几种常用的参数和非参数统计量,用再抽样方法进行了实例研究,其中采用滑动分块自助法以消除自相关性的干扰,该方法为估计水文时间序列的趋势、跳跃的显著性水平给出了新的思路与途径。

This paper was mainly focused on analyzing the trend and the change point in hydrological time series by using the resampling methods. In order to satisfy the characteristics of auto correction, a moving blocks bootstrap was used to analyze the hydrological time series with the conventional trend and change point analysis methods, i.e., the t test, the Mann-Kendall test, the Spearman＇s rho test and the median test for trend, and the runs test, the Mann-Whitney test, the cumulative sum test and the deviation test for change point detection. With the Yellow River as a case study, it＇s shown that the resampling methods are very valuable for the time series analysis.